Abstract

The time delay between signals received at two spatially separated microphones is estimated using the location of the peak in the cross-correlation sequence between the two signals. The estimated delay is usually expressed in integer multiple of the sampling interval. The robustness of the delay estimation is affected by the degradation of the signals due to waveform distortion, noise, reflections, and reverberation. The broadband nature of speech signals is exploited to derive multiple evidence for the time delay from each frequency component of the signal. This helps to reduce the effects of waveform distortion, as the cross-correlation is computed on individual components. The multiple evidence provides robustness in the estimation of the time delay. The accuracy of the estimated time delay depends on the error between the integer delay and the true fractional delay. Methods are proposed to estimate the fractional part of the sampling interval of the delay. Robustness of the proposed method of estimation of the fractional part of the delay is examined for different types and levels of additive noise, as well for reverberation. Experimental results are given for data collected in an actual live room

Abstract

XAlign: Cross-lingual Fact-to-Text Alignment and Generation for Low-Resource Languages Tushar Abhishek, Shivprasad Sagare, Bhavyajeet Singh, Anubhav Sharma, Manish Gupta∗ and Vasudeva Varma Information Retrieval and Extraction Lab, IIIT Hyderabad, India {tushar.abhishek,shivprasad.sagare,bhavyajeet.singh,anubhav.sharma}@research.iiit.ac.in, {manish.gupta,vv}@iiit.ac.in Abstract Multiple critical scenarios (like Wikipedia text generation given English Infoboxes) need au- tomated generation of descriptive text in low resource (LR) languages from English fact triples. Previous work has focused on English fact-to-text (F2T) generation. To the best of our knowledge, there has been no previous at- tempt on cross-lingual alignment or generation for LR languages. Building an effective cross- lingual F2T (XF2T) system requires alignment between English structured facts and LR sen- tences. We propose two unsupervised meth- ods for cross-lingual alignment. We contribute XALIGN, an XF2T dataset with 0.45M pairs across 8 languages, of which 5402 pairs have been manually annotated. We also train strong baseline XF2T generation models on XAlign dataset

Abstract

We consider the problem of achieving fair classification in Federated Learning (FL) under data heterogeneity. Most of the approaches pro- posed for fair classification require diverse data that represent the different demographic groups involved. In contrast, it is common for each client to own data that represents only a single demographic group. Hence the existing approaches cannot be adopted for fair classification models at the client level. To resolve this challenge, we propose several aggregation techniques. We empirically validate these techniques by comparing the resulting fairness and accuracy on CelebA and UTK datasets

Abstract

We consider a budgeted combinatorial multi-armed bandit setting where, in every round, the algorithm selects a super-arm consisting of one or more arms. The goal is to minimize the total expected regret after all rounds within a limited budget. Existing techniques in this literature either fix the budget per round or fix the number of arms pulled in each round. Our setting is more general where based on the remaining budget and remaining number of rounds, the algorithm can decide how many arms to be pulled in each round. First, we propose CBwK-Greedy-UCB algorithm, which uses a greedy technique, CBwK-Greedy, to allocate the arms to the rounds. Next, we propose a reduction of this problem to Bandits with Knapsacks (BwK) with a single pull. With this reduction, we propose CBwK-LP-UCB that uses PrimalDualBwK ingeniously. We rigorously prove regret bounds for CBwK-LP-UCB. We experimentally compare the two algorithms and observe that CBwK-Greedy- UCB performs incrementally better than CBwK-LP-UCB. We also show that for very high budgets, the regret goes to zero.

Abstract

We introduce MUGL, a novel deep neural model for large-scale, diverse generation of single and multi-person pose-based action sequences with locomotion. Our con- trollable approach enables variable-length generations cus- tomizable by action category, across more than 100 cate- gories. To enable intra/inter-category diversity, we model the latent generative space using a Conditional Gaussian Mixture Variational Autoencoder. To enable realistic gen- eration of actions involving locomotion, we decouple local pose and global trajectory components of the action se- quence. We incorporate duration-aware feature represen- tations to enable variable-length sequence generation. We use a hybrid pose sequence representation with 3D pose sequences sourced from videos and 3D Kinect-based se- quences of NTU-RGBD-120. To enable principled com- parison of generation quality, we employ suitably modi- fied strong baselines during evaluation. Although smaller and simpler compared to baselines, MUGL provides better quality generations, paving the way for practical and con- trollable large-scale human action generation.

Abstract

Online Social Media platforms (OSMs) have become an essential source of information. The high speed at which OSM users submit data makes moderation extremely hard. Consequently, besides offering online networking to users, the OSMs have also become carriers for spreading fake news. Knowingly or unknowingly, users circulate fake news on OSMs, adversely affecting an individual’s offline activity. To counter fake news, several dedicated websites (referred to as fact-checkers) have sprung up whose sole purpose is to identify and report fake news incidents. There are well-known datasets of fake news; however, not much work has been done regarding credible datasets of fake news in India. Therefore, we design an automated data collection pipeline to collect fake incidents reported by fact-checkers in this work. We gather 4,803 fake news incidents from June 2016 to December 2019 reported by six popular fact-checking websites in India and make this dataset (FakeNewsIndia) available to the research community. We find 5,031 tweets on Twitter and 866 videos on YouTube mentioned in these 4,803 fake news incidents. Further, we evaluate the impact of fake new incidents on the two prominent OSM platforms, namely, Twitter and YouTube. We use popularity metrics based on engagement rate and likes ratio to measure impact and categorize impact into three levels — low, medium, and high. Our learning models use features extracted from text, images, and videos present in the fake news incident articles written by fact-checking websites. Experiments show that we can predict the impact (popularity) of videos (appearing on fake news incident articles) on YouTube more accurately (with baseline accuracy ranging from 86% to 92%) as compared to the impact (popularity) of tweets on Twitter (with baseline accuracy of 37% to 41%). We need to build more intelligent models that predict tweets’ impact, appearing in fact-checking incident articles on Twitter as future work. 1.

Abstract

Hashtag segmentation is the task of breaking a hashtag into its constituent tokens. Hashtags often encode the essence of usergenerated posts, along with information like topic and sentiment, which are useful in downstream tasks. Hashtags prioritize brevity and are written in unique ways - transliterating and mixing languages, spelling variations, creative named entities. Benchmark datasets used for the hashtag segmentation task - STAN, BOUN - are small in size and extracted from a single set of tweets. However, datasets should reflect the variations in writing styles of hashtags and also account for domain and language specificity, failing which the results will misrepresent model performance. We argue that model performance should be assessed on a wider variety of hashtags, and datasets should be carefully curated. To this end, we propose HashSet, a dataset comprising of: a) 1.9k manually annotated dataset; b) 3.3M loosely supervised dataset. HashSet dataset is sampled from a different set of tweets when compared to existing datasets and provides an alternate distribution of hashtags to build and validate hashtag segmentation models. We show that the performance of SOTA models for Hashtag Segmentation drops substantially on proposed dataset, indicating that the proposed dataset provides an alternate set of hashtags to train and assess models. Datasets and results are released publicly and can be accessed

Abstract

Advancements in adaptive object detection can lead to tremendous improvements in applications like autonomous navigation, as they alleviate the distributional shifts along the detection pipeline. Prior works adopt adversarial learn- ing to align image features at global and local levels, yet the instance-specific misalignment persists. Also, adaptive object detection remains challenging due to visual diver- sity in background scenes and intricate combinations of ob- jects. Motivated by structural importance, we aim to attend prominent instance-specific regions, overcoming the feature misalignment issue. We propose a novel resIduaL seLf- attentive featUre alignMEnt (ILLUME) method for adaptive object detection. ILLUME comprises Self-Attention Fea- ture Map (SAFM) module that enhances structural atten- tion to object-related regions and thereby generates domain invariant features. Our approach significantly reduces the domain distance with the improved feature alignment of the instances. Qualitative results demonstrate the ability of ILLUME to attend important object instances required for alignment. Experimental results on several benchmark datasets show that our method outperforms the existing state-of-the-art approaches.

Abstract

Fully entangled fraction (FEF) is a significant figure of merit for density matrices. In bipartite d ⊗ d quantum systems, the threshold value FEF > 1/d, dictates prominent implications for tasks like teleportation and entanglement distillation. Like separability, the value of FEF is also related to the choice of global basis of the underlying Hilbert space. A state having its FEF ≤ 1/d, might give a value > 1/d in another global basis. A change in the global basis corresponds to a global unitary action on the quantum state. In the present work, we find that there are quantum states whose FEF remains less than 1/d, under the action of any global unitary i.e., any choice of global basis. Based on the fixed spectrum of density matrices, we provide necessary and sufficient criteria in two qubit systems which also prove to be sufficient in any d⊗d dimensions, to identify such states. Further, we prove that states having their FEF bounded by 1/d under any global unitary, form a convex and compact set. This entails the distinction of states whose FEF can be increased beyond 1/d, through unitary action. The demarcation is of paramount importance as it provides for the identification of states which can prove to be useful in teleportation and entanglement distillation after the action of global unitary operators, as underpinned by illustrations in our work.

Abstract

n this paper, we develop a robust 3D garment digitiza- tion solution that can generalize well on real-world fashion catalog images with cloth texture occlusions and large body pose variations. We assumed fixed topology parametric template mesh models for known types of garments (e.g., T- shirts, Trousers) and perform mapping of high-quality tex- ture from an input catalog image to UV map panels corre- sponding to the parametric mesh model of the garment. We achieve this by first predicting a sparse set of 2D landmarks on the boundary of the garments. Subsequently, we use these landmarks to perform Thin-Plate-Spline-based tex- ture transfer on UV map panels. Subsequently, we employ a deep texture inpainting network to fill the large holes (dueto view variations & self-occlusions) in TPS output to gen- erate consistent UV maps. Furthermore, to train the su- pervised deep networks for landmark prediction & texture inpainting tasks, we generated a large set of synthetic data with varying texture and lighting imaged from various views with the human present in a wide variety of poses. Addition- ally, we manually annotated a small set of fashion catalog images crawled from online fashion e-commerce platforms to finetune. We conduct thorough empirical evaluations and show impressive qualitative results of our proposed 3D gar- ment texture solution on fashion catalog images. Such 3D garment digitization helps us solve the challenging task of enabling 3D Virtual Try-on